Characters · Lesson 6 of 6
How to train a LoRA on your face or style
Teach the AI your face or your style from one photo or a folder of images, then use it in any prompt.
You need: One photo, or 10 to 20 photos of the same face or style.
Step by step
Choose a training route
Fast Trainer runs in about two minutes. Zip upload brings an image collection you already have. You can also upload up to 30 single images.
Start from one photo if that is all you have
The single-image option makes 11 portrait variations from one photo and trains on those.
Or upload 10 to 20 images
Captions are written for you. Variety wins: different angles, light and backgrounds teach the model what stays the same.
Find it in your Library
Open Library, then LoRAs. Each one shows its trigger word and a Create button.
Use it in a prompt
In Qwen Image 2.1, open Add LoRA under the prompt, browse your LoRAs and pick yours. Then write the scene.
Why it works
- Mixed angles and lighting make a flexible LoRA. Twenty near-identical selfies make a stiff one.
- A LoRA is for a look you reuse constantly. For a recurring cast in films, a saved character is faster.
Your practice task
Train a style LoRA from 15 of your own photos and use it on a subject you never shot.
Read the transcript(1:26)
Teach the AI your face, or your style, then bring your own LoRA into prompts across PixelDojo Start in the LoRA Trainer, and choose your training route Fast Trainer runs in about 2 minutes, while Zip file Upload lets you bring an existing image collection You can also upload up to 30 individual images, or start with 1 image in Character Studio to build a training set For this walkthrough, choose Fast Trainer The Single Image option takes 1 photo, and generates 11 portrait variations for training That gives you a way to build your portrait model from just 1 starting image Already have your training images? Choose Zip file for a collection of 10 to 20 images Captions are added automatically, so you do not have to write each one yourself Once training finishes, open Library, and select LoRAs Your trained models live here, with their trigger words and create buttons, ready to take into an image tool Now bring your training into a prompt Here, in Qwen Image 2.1, open Add LoRA beneath the prompt box to access your saved models Choose Browse Your LoRAs, then pick your trained model This is where you connect your personal face or style with the prompt you want to create Choose your images, train your LoRA, and reuse it Next, try your face or style in a new scene